Key Highlights
Up to 20 PFLOPS
High-performance AI computing for demanding training and inference workloads.
748GB Coherent
A large unified CPU-GPU memory architecture for working with large AI models.
ConnectX-8
High-bandwidth QSFP networking for advanced AI infrastructure and scalable deployments.
NVIDIA AI Stack
Ubuntu with NVIDIA AI Developer Tools for local AI development and deployment.
Introduction
The ASUS ExpertCenter Pro ET900N G3 AI Supercomputer is designed for organisations, AI developers, researchers, and data scientists who need substantial AI compute capacity close to where they work. Built on NVIDIA DGX Station architecture, it combines a 72-core NVIDIA Grace CPU and NVIDIA Blackwell Ultra GPU in a deskside tower system.
Its 748GB coherent memory architecture is designed to help developers work with large models and memory-intensive workloads locally. With up to 20 PFLOPS of AI performance, the platform is suited to demanding AI development, model inference, deep learning research, and generative AI workflows.
The system also integrates NVIDIA AI software tools and high-speed networking to support professional AI environments. It is a specialised AI computing platform rather than a conventional consumer desktop PC.
Features and Technologies
NVIDIA GB300 Grace Blackwell Ultra
Combines NVIDIA Grace CPU technology and a Blackwell Ultra GPU through NVIDIA NVLink-C2C to enable high-bandwidth, coherent communication within the superchip.
748GB Coherent Memory
Combines 496GB of LPDDR5X CPU memory and 252GB of HBM3e GPU memory into a large coherent memory architecture for AI workloads.
NVIDIA ConnectX-8 SuperNIC
Provides two QSFP112 networking ports for high-speed connectivity in advanced AI infrastructure. Compatible transceivers and network equipment may be required.
NVIDIA AI Software Stack
Ubuntu with NVIDIA AI Developer Tools provides a software foundation for AI development, model inference, data science, and accelerated computing.
PCIe Expansion
Three PCIe Gen 5 expansion slots provide options for compatible add-in hardware, including supported NVIDIA RTX PRO Blackwell workstation graphics cards.
Enterprise Management
An ASPEED AST2600-based BMC, IPMI 2.0, DMTF Redfish API support, and onboard TPM 2.0 support enterprise system management and security workflows.
Large-Model AI Computing at the Deskside
The combination of NVIDIA GB300 compute, 748GB coherent memory, NVLink-C2C, and high-speed ConnectX-8 networking is designed for large-scale AI development and inference without relying exclusively on remote cloud compute. Actual performance depends on the workload, software, and configuration.
Technical Specifications
Specifications and supported configurations may vary by region. Optional expansion hardware, transceivers, Wi-Fi module, and antenna may need to be purchased separately. Confirm the final configuration with the supplier before ordering.
Ideal Use Cases
Large Language Models
Local LLM inference, model experimentation, fine-tuning and development workflows that benefit from a large coherent memory pool.
Generative and Agentic AI
Develop and deploy generative AI applications, autonomous AI agents and multi-step AI workflows in a controlled local environment.
Research and Data Science
Support deep learning research, accelerated computing, data science and demanding AI experimentation.
Enterprise AI Infrastructure
Bring AI compute closer to teams that need local data control, high-bandwidth networking and managed infrastructure.
Before You Buy / Install
- Confirm the exact regional configuration, warranty coverage, delivery details and included accessories with the supplier.
- Plan for a suitable power connection. The system is specified with a 1600W Titanium ATX power supply and a global input range of 200–240V AC.
- Allow for the system's 27kg net weight and tower dimensions of 584 x 232 x 565mm when planning placement and installation.
- The optional PCIe graphics card, Wi-Fi module, antenna and QSFP transceivers are not necessarily included in the base configuration. Confirm what is supplied.
- The listed storage configuration includes two 2TB NVMe SSDs. Verify drive population, RAID configuration and any additional storage requirements before deployment.
- For sustained AI workloads, maintain suitable ventilation and observe the manufacturer's recommended operating temperature range of 10°C to 35°C.
Frequently Asked Questions
What is the ASUS ExpertCenter Pro ET900N G3?
It is a deskside AI supercomputer built on NVIDIA DGX Station architecture. It combines the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, 748GB coherent memory and NVIDIA AI software tools for demanding local AI workloads.
How much AI performance does the ET900N G3 deliver?
ASUS specifies up to 20 PFLOPS of AI performance. Actual results depend on the workload, model, software stack and operating conditions.
How much memory does this AI supercomputer have?
The listed system has 748GB of coherent memory, comprising 496GB LPDDR5X CPU memory and 252GB HBM3e GPU memory.
Which operating system comes with the ET900N G3?
The system is specified with Ubuntu and NVIDIA AI Developer Tools. Confirm the supplied software configuration with the seller before purchase.
Can additional storage be installed?
The product listing describes four M.2 2280 storage slots. The listed configuration includes two 2TB NVMe M.2 SSDs. Check the supported drive type, slot configuration and installation requirements before adding drives.
Does it include an additional NVIDIA RTX PRO graphics card?
The listing identifies compatible NVIDIA RTX PRO Blackwell graphics cards as optional. Do not assume an additional RTX PRO card is included unless it is explicitly stated in the final quotation.
What power supply does the system use?
The listed configuration specifies a 1600W Titanium ATX power supply. The manufacturer notes that maximum power output can depend on regional input voltage and the power-cord assembly.
Who should consider the ASUS ExpertCenter Pro ET900N G3?
It is intended for AI developers, researchers, data scientists and enterprises requiring substantial local AI compute for model development, inference, generative AI and related workloads.
Build Your PC with Confidence
Looking for the right components for your next workstation or AI computing setup? Explore Clarion Computers for processors, graphics cards, motherboards and essential PC components to support your build.